Ever since OpenAI introduced ChatGPT in November 2022, the world has seen the launch of a torrent of Large Language Models (LLMs) that can give human-like answers by being very proficient in understanding and using language. Their ability to produce a meaningful answer to virtually any question in no time has wowed the world.
Unlike in the past, where technological advances impacted a few areas of the economy, LLMs have impacted virtually every single aspect of the economy. They have been disrupting the traditional models everywhere, posing challenges at different levels — from plagiarism to loss of jobs and how to access scarce and prohibitively expensive GPUs that are at the heart of LLMs.
The LLM space is evolving so rapidly that every other day, we hear about a novel or more powerful solution hitting the Internet.
They also raise questions in connection with whether India should develop its own LLM or focus on developing applications based on the existing LLMs instead of reinventing the wheel, wasting meagre financial and computational resources.
The two divides
India faces a still bigger challenge. These LLMs can pose a serious challenge to the digital divide, which is already wide due to various reasons. The divide being caused by LLMs is happening at two levels.
The primary divide between those who can access the digital world and enjoy the benefits, and those who cannot, is getting deeper and more complicated. This is because the benefits of LLMs are currently being enjoyed by a section of the population. Though there are free versions of LLMs such as ChatGPT, Google’s Geminiare, Facebook’s Meta AI, Elon Musk’s Grok, and Chinese LLM DeepSeekR1, only a few sections of people are using them. There is little awareness among the people regarding what they are all about, and how they can benefit from them.
Yet, India has emerged as a top user of these services. The fact that OpenAI already announced that India has emerged as their second biggest market in terms of the number of users for them reflects how fast the usage is growing.
This has opened up a range of contrasts at multiple levels — among the general population, employees (private and public), and students. There is a significant gender disparity in LLM access and usage, with women being grossly under-represented.
A section of students at different levels has already started using LLMs for research and building applications; the majority of students are nowhere near using them. This puts the latter category at a disadvantage at a time when the youth are finding it difficult to find jobs.
Affordability angle
While this is about the general digital divide, there is another divide that is sharply emerging — among the digital literates. Though they know what ChatGPT is, only a few can afford to go for the paid subscriptions that give them a lot more features, way higher than those available in the free versions.
Subscription factor
Those who have access to the paid subscriptions have become more productive and can learn and earn more, while others are falling behind.
The subscription rate cards, which hover around ₹2,000 a month, might look like peanuts in the West, but it is a significant amount even for those who are earning ₹50,000 a month. This leaves a large number of digitally literate Indians out of the LLM party.
There is another factor that is hindering proper use: the lack of awareness on how to elicit the best output from LLM.
Unlike in traditional searches where one can use ‘keywords’ to get answers, LLMs require ‘prompts’ to help them understand what the user wants to ask. One needs to know a little bit of ‘prompt engineering’ (a set of best possible suggestions).
Since all the existing LLMs are ‘trained’ in the West, there is a lot of scope for bias in the outputs that they produce. They may not be in a position to understand the cultural, historical, social, political, literary, and geographical nuances of India. This could lead to skewed, uninformed, and biased output. This, in turn, could negatively or wrongly influence the users in India about the issues related to their own country.
As the LLM space is evolving so rapidly, it is about time for the country to think about these issues seriously and find solutions, as it can exclude a sizeable chunk out of this technological revolution.
Forming a Joint Parliamentary Committee or a panel with AI experts and other stakeholders to suggest ways to overcome this challenge would help.